The full spectrum of clinical stages of psychosis among mentally ill prisoners in New South Wales (NSW), Australia

Abstract

Research on the association between psychosis and criminal offending has typically focused on violent offenders with chronic psychotic illness. This stages of psychosis in prison (SOPP) study used a clinical staging approach to identify adult men referred to prison mental health services who had an at-risk mental state (ARMS), first episode of psychosis (FEP) or an established psychotic illness. Of the 105 participants included, 6% were determined to have FEP, 6% met ARMS criteria and the remainder had an established psychotic illness. Compared to a prison control sample, individuals on the psychosis spectrum were found to have higher levels of social disadvantage and other co-occurring mental health and substance use problems but were not more likely to have committed a violent offence. These findings support the notion that risk of criminal justice contact and complex illness burden exist across the full spectrum of psychotic illness.

Introduction

The over-representation of people with severe mental illness in contact with the criminal justice system is well established (Brugha et al., 2005; Fazel & Danesh, 2002; Fazel et al., 2016; Teplin, 1984; Yee et al., 2020). A systematic review involving 33,588 adult prisoners from 109 samples worldwide (Fazel & Seewald, 2012) has estimated the pooled prevalence of psychosis to be 3.6% in males (95% CI = 3.1–4.2) and 3.9% in females (95% CI = 2.7–5.0), almost 8 times higher than the 0.5% prevalence rate typically found in general population samples (Saha et al., 2005). Most research on psychosis and offending examines serious violence, especially homicide, even though psychotic illness is associated with an increased likelihood of most forms of offending (Yee et al., 2020).

Until recently, most studies have focused on the presence or absence of a psychotic disorder when seeking to understand the relationship between psychosis and offending. However, outside of justice settings, clinical staging models of psychosis – with a focus on identification and intervention as early as possible in the course of illness – are well established (McGorry et al., 2006, 2007; Yung & McGorry, 1996). The dominant clinical staging model of psychosis adopts a dimensional perspective whereby psychosis is seen to progress along a continuum, from vulnerability to developing the illness at one end through to chronic and treatment-resistant illness at the other. Although there is no assumption that an individual will inevitably progress along the continuum in a linear fashion, different stages of illness may be associated with different characteristics and treatment needs (McGorry et al., 2006, 2007).

Hence, the relationship between psychosis and criminal justice contact may also differ across stages of illness.

Emerging evidence suggests that the earliest stages of psychosis, when positive psychotic symptoms and lack of insight are often prominent, may present a period of particular risk of violence and other offending (Nielssen et al., 2012; Yee et al., 2011), whilst those in the chronic stages of illness when negative symptoms can emerge (e.g. social withdrawal, blunted affect and poor abstract thinking) may exhibit a lower risk (Swanson et al., 2006). The first episode of psychosis (FEP), typically defined as the period between the onset of frank psychotic symptoms and the start of antipsychotic treatment (Register-Brown & Hong, 2014), has been demonstrated to be a period associated with a high risk of harm to both self and others, including risk of violence (Challis et al., 2013; Large & Nielssen, 2011; Latalova, 2014). A meta-analysis examining the spectrum of violent offending during the FEP found that 35.0% of patients in their FEP commit some form of violence, 16.6% commit serious violence and 0.6% commit severe violence (Large & Nielssen, 2011). High rates of aggression have also been reported among FEP patients in the community. For example, a United Kingdom (UK) study found that approximately 1 in 7 individuals had been arrested or charged for violent offences within the first 12 months after contact with early psychosis intervention services (Whiting et al., 2020); another study found that two thirds of patients with a schizophrenia spectrum disorder and a history of violent offending had committed their first violent offence prior to being diagnosed and treated (Hachtel et al., 2018).

Whereas the period preceding the FEP, namely the ‘at-risk mental state’ (ARMS; see Purcell et al., 2015), has received considerable recent attention on the basis that it may be possible to identify and intervene with individuals at increased risk of later psychosis (Fusar-Poli et al., 2012; Yung et al., 2004; Yung & McGorry, 1996), only two previous studies have attempted to investigate the ARMS in incarcerated samples. In one Irish study of young offenders in detention, 23% of new committals met the ultra high risk (UHR) criteria using the Comprehensive Assessment of At-Risk Mental State (CAARMS) instrument (Flynn et al., 2012). In another study examining early psychosis within an adult male prison population in London, 5% were found to meet the UHR criteria and another 3% met the FEP threshold (Jarrett et al., 2012; Jarrett, Jamieson-Craig et al., 2016; Jarrett, Valmaggia et al., 2016). The UHR is also associated with an increased risk of non-violent criminal offending in non-prison samples. In a longitudinal cohort study of 711 young Australians aged 12 to 25 years presenting to youth mental health services in Victoria, those at-risk for psychosis (n = 271; 38%) had significantly higher rates of non-violent criminal charges (OR = 1.98, 95% CI = 1.14–3.43, p = .015), any criminal convictions (OR = 2.20, 95% CI = 1.05–4.60, p = .037) and non-violent convictions (OR = 2.88, 95% CI = 1.12–7.40, p = .028) compared to not-at-risk controls (Purcell et al., 2015).

The temporal relationship between age of offending and age of illness onset has been postulated to differentiate subgroups of offenders with schizophrenia spectrum disorders (Kooyman et al., 2012), with the majority of mentally ill offenders thought to begin their antisocial trajectory after illness onset (Hodgins, 2008). Although it is difficult to establish with certainty the temporal contiguity between psychosis and violence (Douglas et al., 2009), the association is strongest when the temporal contiguity is within 12 months (Van Dorn et al., 2012). This suggests that examining the timing of onset of both offending and illness is important to better understand the association between psychosis and risk of criminal offending as well.

This stages of psychosis in prison (SOPP) study aims to identify individuals experiencing psychosis at any point along the clinical spectrum – from early psychosis (including the ARMS) to chronic or established psychosis – amongst adult men referred to prison mental health services in New South Wales (NSW), Australia, and to compare this group to a non-psychotic prison control sample.

Method

Participants and procedure: psychotic-spectrum sample

The participants consist of male prisoners aged 18 to 64 years who entered the NSW Metropolitan Reception and Remand Centre (MRRC) between 1 February 2015 and 31 January 2016 and were referred following standard prison-reception health screening to prison mental health services due to concerns about possible psychotic illness. In NSW, primary health nurses undertake standard reception screening that includes a general mental health assessment incorporating a series of questions about psychiatric history, current symptoms of mental illness and self-harm history and/or ideation. This brief screening assessment is then used by these generalist nurses to inform referrals to prison mental health services. The MRRC is the largest reception and remand centre for adult men in NSW and the largest prison in Australia. During the 12-month study recruitment period, a total of 5976 male prisoners were received into the MRRC and, of these, 2185 (36.6%) were referred to mental health services.

Referrals for the study were sought from mental health clinicians (psychiatrists and mental health nurses) who had been instructed to identify men referred to their service with any indication of possible psychosis or experiences which might suggest that they were at risk of developing a psychotic illness. The clinicians did not use a psychosis-specific screener or questionnaire for this, rather making referrals on the basis of clinical judgement. Those with a recorded diagnosis of an organic brain disorder, a recorded intellectual disability (i.e. IQ < 70), a lack of proficiency in spoken English sufficient for interview, a lack of capacity to consent or who were deemed to present too great a risk of harm to others to be safely interviewed were excluded. Clinician-based identification of potential participants was augmented with regular reviews of electronic mental health referral records in order to ensure that potential participants were not missed, as well as to check the eligibility of the referrals made by the mental health clinicians.

Eligible individuals referred by custodial mental health clinicians or identified following regular review of electronic referral records were approached by the researchers about participating in the study. In view of the likelihood that due to restrictions on prisoner access not all eligible individuals would be readily accessible, the approaches were made in a randomised order. A random sequence generator (http://www.randomizer.org) was used to determine the order in which the eligible individuals were approached, and those who were missed in any given week due to time constraints or custodial procedural restrictions were added to the eligible-to-approach list for the following week. All eligible individuals approached were given an information sheet about the study, and written consent was sought from all those who agreed to participate. Participants were reimbursed with AUD$10 post-participation (a standard amount for research in prison settings) which was paid into their prison account.

Instruments: psychotic-spectrum sample

Each participant completed a two-hour face-to-face interview. In addition to this, participants’ health records from Justice Health and Forensic Mental Health Network files and criminal records from the Corrective Services New South Wales (CSNSW) Offender Integrated Management System (OIMS) were examined to obtain corroborative mental health and offending information.

A structured questionnaire was employed to obtain key sociodemographic information (e.g. age, birth country, marital status, legal status, employment status and Aboriginal and Torres Straits Islander status), clinical information (e.g. history of psychiatric diagnoses and treatments, self-harm and substance use) and offending information (e.g. offence type, victim type and weapon use). Age-related variables were also included as part of the questionnaire; participants were asked to indicate age of onset for the following key mental health and offending events: first illicit drug use, first becoming mentally unwell (feeling that ‘something is not right’), first mental health contact, first mental health treatment, first mental health diagnosis, first mental health admission, first police contact, first criminal charge, first custody and age at time of index offence. To determine whether or not participants had a prior history of psychosis, they were asked ‘have you ever been diagnosed with a psychotic illness?’, and for those who reported receiving antipsychotic treatment in the past, whether or not this treatment was for psychosis.

The CAARMS (Yung et al., 2005), a semi-structured interview schedule used to assess UHR status, was administered to all interviewed participants. It includes seven subscales: Positive Symptoms, Cognitive Change (Attention/Concentration), Emotional Disturbance, Negative Symptoms, Behavioural Changes, Motor/Physical Changes and General Psychopathology. Each subscale is scored from 0 (absent) to 6 (psychotic/severe) and measures threshold, frequency, duration of symptoms, level of distress caused by symptoms and whether or not symptoms occurred in the context of substance use. In addition to its use in two prison samples (Flynn et al., 2012; Jarrett et al., 2012; Jarrett, Valmaggia et al., 2016), the CAARMS has been used to identify the ARMS among a sample of adult Aboriginal prisoners in NSW and was found to be effective in differentiating between those diagnosed with psychotic illness and healthy controls (Hamilton, 2008). To meet UHR criteria, only the Positive Symptoms subscale (encompassing Unusual Thought Content, Non-Bizarre Ideas, Perceptual Abnormalities and Disorganised Speech) was used (Flynn et al., 2012) and the three types of UHR were established: (i) the Vulnerable group, having a first-degree relative with psychosis or having schizotypal personality disorder; (ii) the Attenuated Psychosis group, having psychotic symptoms of subthreshold intensity or frequency; and (iii) the Brief Limited Intermittent Psychotic Symptoms (BLIPS) group, having frank psychotic symptoms that spontaneously resolve within seven days. Those meeting UHR criteria were also required to have experienced a functional decline, as measured by the Social and Occupational Functioning Scale (SOFAS) attached to the CAARMS (Yung et al., 2005). Additionally, the CAARMS was also utilised for the identification of those meeting the threshold for a current psychotic episode, defined as the presence of frank psychotic symptoms lasting longer than one week.

The Structured Clinical Interview for DSM-IV-TR Axis-II Personality Disorders (SCID-II; First et al., 1997) – a semi-structured interview-based instrument used to diagnose 10 personality disorders according to Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM-IV-TR; American Psychiatric Association, 2000) criteria that has been tested in prison settings (Ullrich et al., 2008; Warren et al., 2002) – was administered to determine the presence of schizotypal, schizoid, paranoid and antisocial personality disorders. The schizotypal module was also used to determine whether or not participants met the Vulnerable group criterion on the CAARMS.

Participants, instruments and procedure: non-psychotic control group

To investigate whether or not prisoners with a psychosis-spectrum illness or who are at-risk for psychosis differ from non-psychotic prisoners in NSW in terms of sociodemographic, clinical and offending characteristics, a control group was sourced from a contemporaneous prison health survey – the 2015 Network Patient Health Survey (NPHS; Justice Health and Forensic Mental Health Network, 2017). This cross-sectional health survey of 1132 adult prisoners was conducted in 2015 across many NSW prisons. In a face-to-face survey interview, participants were asked a range of sociodemographic, offending, mental health, alcohol and drug use and physical health questions. A list of comparable sociodemographic, offending and clinical questions for both SOPP cases and NPHS sample can be found under Supplementary Material. To obtain the control sample for the present study, the male participants of the NPHS survey who reported a lifetime diagnosis of a psychotic disorder (when asked ‘have you ever been told by any clinician or doctor that you have schizophrenia or a psychotic disorder, including drug-induced psychosis?’) or who screened positive on at least one symptom dimension (thought insertion, paranoia, strange experiences and hallucinations) of the Psychosis Screening Questionnaire (PSQ; Bebbington & Nayani, 1995) were excluded. The remaining 471 non-psychotic male participants (out of 757 total) were retained to form the control group for the present study.

Statistical analysis

In order to establish the prevalence of the full spectrum of clinical stages of psychosis in the SOPP sample (cases), the use and interpretation of the CAARMS was extended to enable the grouping of participants with a prior history of psychosis according to their current level of psychosis symptomatology. Individuals with established psychosis were categorised based on whether or not they displayed any current psychosis-threshold symptoms or subthreshold symptoms. After those with no symptoms of psychosis (negative CAARMS and no lifetime history of psychosis) had been excluded, the following five clinical-stage groups were established:

1. UHR (CAARMS subthreshold symptoms, no lifetime psychosis history);

2. FEP (CAARMS threshold for psychosis met, no lifetime psychosis history);

3. Established psychosis currently in remission (negative CAARMS, presence of lifetime psychosis history);

4. Established psychosis with residual symptoms (CAARMS subthreshold symptoms, presence of lifetime psychosis history);

5. Established psychosis with current threshold-level symptoms of psychosis (CAARMS threshold for psychosis met, presence of lifetime psychosis history).

Given the relatively small size of the five groups, comparative analyses of sociodemographic, clinical and offending characteristics were undertaken across three larger groups defined by the current level of psychosis symptomatology: Group 1 – No Current Psychosis (remitted); Group 2 – Subthreshold Psychosis (residual or at-risk); Group 3 – Psychosis Threshold Met (active psychosis, either unremitted or first episode). Age-related variables for key clinical and offending characteristics are presented in a descriptive manner for the psychosis-spectrum group as a whole.

Since the non-psychotic prison control group was obtained from a separate study in which there was a deliberate over-sampling of Aboriginal and/or Torres Strait Islander participants, the weighting that was applied in that original NPHS study to take account of this over-sampling was retained in all analyses involving the control group in the present study (i.e. no additional weighting method was introduced). The control group therefore comprises a total weighted sample size of 489. Further details of the over-sampling and weighting method applied in the NPHS study to the Aboriginal and/or Torres Strait Islander participants can be found in the 2015 Network Patient Health Survey Report (Justice Health and Forensic Mental Health Network, 2017).

Chi-square analyses were used to compare categorical dependent variables, logistic regression (univariate and multivariate) analyses were used to test the differences between cases and controls for categorical dependent variables and Mann–Whitney U tests were used for continuous dependent variables (selected due to the non-parametric nature of the data). Any subgroup findings with five or fewer individuals are not reported in order to avoid potential re-identification. All analyses were performed using SPSS Statistics for Windows, v25.0 (IBM, 2017), with p < .05 denoting statistical significance.

Results

The flow diagram of the participants recruited and included in the SOPP sample is shown in Figure 1. Of the 2185 individuals who were referred to in-reach mental health services during the study’s 12-month recruitment period, 306 were referred to the study (14.0%), of which 280 were deemed eligible to participate (91.5%). A total of 166 of these 280 individuals were randomly approached, of which 55 refused to participate (33.1%), 1 was further excluded because the confirmed reception date was outside the study recruitment period and 3 were excluded due to incomplete CAARMS assessments, yielding a total of 107 participants. However, 2 participants were further excluded from the analyses due to the absence of any current or prior evidence of psychosis or at-risk for psychosis status, resulting in a final sample of 105 psychosis-spectrum cases for the analyses.

Figure 1.

Flow chart of the recruitment procedure for the psychosis-spectrum sample.

Prevalence of the clinical stages of psychosis

Five psychosis-spectrum subgroups were identified based on the participants’ CAARMS profiles and lifetime histories of psychosis (Figure 2). Of the 12 participants who reported no previous history of psychosis (11.4%), 6 met UHR criteria (5.7%) and 6 met criteria for FEP (5.7%); the remaining 93 participants reported a lifetime history of psychosis and thus were regarded as being in the established stage of psychosis (88.6%). Of these 93 participants, 22 were found to be in remission (21.0%), 30 were displaying residual symptoms (28.6%) and 41 met the psychosis threshold on the CAARMS and remained actively unwell (39.0%). The number of individuals in the three UHR subtypes is not reported due to their small number. The sociodemographic, clinical and offending characteristics for the three psychosis-spectrum subgroups (n = 105) are presented in Table 1, along with the results of the chi-square analysis undertaken to compare the groups.

Figure 2.

Prevalence of the different stages of psychosis according to lifetime history and CAARMS profile (n = 105).

Table 1.

Sociodemographic, clinical and offending characteristics of the psychosis-spectrum participants according to psychosis symptomatology at time of interview (n = 105).

Characteristic	Total	No current psychosis (n = 22)	Subthreshold psychosis (n = 36)	Psychosis threshold met (n = 47)
Sociodemographic
Age* (years), M (SD) (range = 18–64)	34 (9.3)	37 (8.6)	31 (8.3)	35 (9.8)
On remand, n (%)	66 (62.9)	12 (54.5)	26 (72.2)	28 (59.6)
First time in custody, n (%)	32 (30.5)	9 (40.9)	11 (30.6)	12 (25.5)
Single, n (%)	66 (62.9)	12 (54.5)	26 (72.2)	28 (59.6)
Indigenous Australian, n (%)	22 (21.0)	5 (22.7)	8 (22.2)	9 (19.1)
Born in Australia, n (%)	80 (76.2)	18 (81.8)	29 (80.6)	33 (70.2)
Unemployed, n (%)	82 (78.1)	19 (90.5)	29 (80.6)	34 (73.9)
<Year 10 education, n (%)	45 (42.9)	8 (36.4)	17 (47.2)	20 (42.6)
Childhood trauma, n (%)	75 (71.4)	13 (59.1)	25 (69.4)	37 (82.2)
Clinical
Family history of mental illness*, n (%)	46 (43.8)	6 (30.0)	13 (36.1)	27 (62.8)
Lifetime depression (self-report), n (%)	65 (61.9)	13 (59.1)	22 (61.1)	30 (65.2)
Lifetime anxiety (self-report), n (%)	56 (53.3)	11 (50.0)	17 (47.2)	28 (59.6)
Lifetime ADHD/ADD (self-report), n (%)	27 (25.7)	6 (27.3)	8 (22.2)	13 (28.3)
Previous MH treatment, n (%)	101 (96.2)	21 (95.5)	34 (94.4)	46 (97.9)
Previous MH admission, n (%)	86 (81.9)	18 (20.9)	31 (36.0)	37 (43.0)
Head injury, n (%)	53 (50.5)	8 (36.4)	20 (55.6)	25 (53.2)
Previous self-harm, n (%)	61 (58.1)	10 (45.5)	21 (58.3)	30 (63.8)
Alcohol use problems, n (%)	74 (70.5)	17 (77.3)	24 (66.7)	33 (70.2)
Illicit drug use*, n (%)	103 (98.1)	20 (90.9)	36 (100.0)	47 (100.0)
Intoxicated at time of offence, n (%)	62 (59.0)	12 (57.1)	24 (66.7)	26 (56.5)
Age (years) alcohol use, M (SD)	15 (4.0)	16 (2.2)	16 (3.8)	15 (4.9)
Age (years) illicit drug use, M (SD)	16 (6.8)	17 (7.6)	15 (5.8)	17 (7.3)
Offending
Violent index offence, n (%)	48 (45.7)	13 (59.1)	13 (37.1)	22 (46.8)
Age (years) at time of offence, M (SD)	34 (9.3)	36 (8.7)	31 (8.4)	35 (9.8)
Victim involved, n (%)	54 (51.4)	15 (68.2)	13 (37.1)	26 (56.5)
Auditory hallucinations at offence, n (%)	51 (48.6)	7 (31.8)	20 (55.6)	24 (51.1)
Command hallucinations at offence, n (%)	33 (31.4)	5 (22.7)	12 (33.3)	16 (35.6)
Delusions at offence, n (%)	56 (53.3)	8 (36.4)	23 (63.9)	25 (55.6)
Treatment non-adherence, n (%)	56 (53.3)	12 (57.1)	16 (55.2)	28 (70.0)

Note. *p < .05. ADHD/ADD = attentional deficit hyperactivity disorder/attention deficit disorder; MH = mental health.

Sociodemographic characteristics of the psychosis-spectrum participants

The psychosis-spectrum participants had a mean age of 33.8 years (SD = 9.3 years) and most were on remand at the time of the interview (62.9%). The majority were born in Australia (76.2%), single (62.9%) and unemployed at the time of the index offence (78.1%). Almost three quarters reported a history of adverse childhood events such as physical abuse, sexual abuse or emotional neglect (71.4%). One in five participants identified as being of Aboriginal or Torres Strait Islander background (i.e. Indigenous Australians; 21.0%).

The chi-square analysis revealed no significant sociodemographic differences among the three groups based on current level of symptomatology except for age at time of interview. Not surprisingly, those with subthreshold symptoms (i.e. UHR or residual psychosis) were found to be younger than those with active psychosis or no current symptoms, F (2, 104) = 3.29, p = .04.

Clinical characteristics of the psychosis-spectrum participants

Just under half of the psychosis-spectrum participants reported having a family history of mental illness (43.8%). Self-reported lifetime comorbid psychiatric diagnoses (in addition to a psychotic illness) were prevalent, with depressive disorder being the most common (61.9%), and almost all participants reported receiving psychiatric treatment in the past for their mental health conditions (96.2%). Over half of the participants reported a history of deliberate self-harm including suicide attempts (58.1%) and, separately, being intoxicated at the time of the index offence (59.0%). A history of problems with alcohol use (70.5%) and illicit drug use (98.1%) was highly prevalent, and mid-adolescence was the period when participants typically started to engage in problematic alcohol use (M = 15.5 years, SD = 4.0 years) and illicit drug use (M = 16.4 years, SD = 6.8 years).

The three groups differ significantly with regard to having a family history of mental illness and a previous history of illicit drug use. Those with threshold-level psychotic symptoms were most likely to have a family history of mental illness, χ²(2) = 8.39, p = .015, whereas a history of illicit drug use was least prevalent in the group without current psychosis, χ²(2) = 7.70, p = .02.

Offending characteristics of the psychosis-spectrum participants

The offence that led to the current period of incarceration was of a violent nature for just under half of the psychosis-spectrum sample (45.7%). Several participants reported experiencing psychotic symptoms at the time of the index offence, including auditory hallucinations (48.6%), command hallucinations (31.4%), holding beliefs that they recognised to be delusional (53.3%) and referential ideas (25.7%). For the 101 participants who reported receiving psychiatric treatment around the time of the offence, about half admitted to being non-adherent to prescribed medications (53.3%). The chi-square analysis revealed no significant statistical differences among the groups with regard to offending characteristics.

Timing of key health and justice events among the psychosis-spectrum participants

Figure 3 displays the mean ages associated with key lifetime mental health and offending events for the psychosis-spectrum sample (n = 105). The average ages of onset for illicit substance use, at-risk illness symptoms (when they first felt something was ‘not right’) and criminal justice system contact coincide at around 16 years (M = 16.0 years for first police contact, 16.4 years for starting to use illicit substances and 16.8 years for onset of mental health problems). Although the first criminal charge occurred soon after initial contact with the police (M = 17.4 years), the participants were older across the remaining key mental health events, including first contact with mental health services and commencement of psychiatric treatment (Ms = 21.3 years). Those needing psychiatric hospital admission (n = 86) typically had their first admission in their mid-twenties (M = 24.8 years). These mental health events, except for psychiatric admission, occurred before the participants’ first imprisonment (M = 22.8 years). The index offence leading to the current incarceration occurred almost 10 years after the first incarceration on average (M = 33.6 years).

Figure 3.

Mean ages for key mental health and offending events for the psychosis-spectrum sample.

Comparing the psychosis-spectrum and non-psychotic participants

Table 2 compares the sociodemographic characteristics of the psychosis-spectrum cases and the non-psychotic controls. The participants on the psychosis spectrum were slightly younger than those without psychosis (Mdn = 33.0 years, SD = 9.2 years vs. Mdn = 36.0 years, SD = 13.1 years) but the difference is not significant (U = 21,988, p > .05). The psychosis-spectrum group was significantly more likely to have been on remand at the time of the interview (OR = 8.17, 95% CI = 5.16–12.96, p < .001), previously incarcerated as an adult (OR = 1.64, 95% CI = 1.04–2.58, p = .032) and single and unmarried at the time of the interview (OR = 1.57, 95% CI = 1.02–2.42, p = .041). Those with psychosis were also significantly more likely to have had lower educational attainment (OR = 3.53, 95% CI = 2.25–5.55, p < .001) and to have been unemployed at the time of the index offence (OR = 6.35, 95% CI = 3.79–10.65, p < .001).

Table 2.

Univariate and multivariate analyses for sociodemographic characteristics for the psychosis-spectrum cases (n = 105) and controls (unweighted n = 471, weighted n = 489).

Variable	Psychosis- spectrum cases	Non-psychotic controls	Statistics
Age (years), Mdn (SD)	33 (9.3)	36 (13.1)	U = 21,988, p > .05
On remand, n (%)	66 (62.9)	83 (17.1)	OR = 8.17 (95% CI = 5.16–12.96)**
Previous custody, n (%)	73 (69.5)	283 (58.2)	OR = 1.64 (95% CI = 1.04–2.58)*
Single status, n (%)	66 (62.9)	252 (52.0)	OR = 1.57 (95% CI = 1.02–2.42)*
Indigenous Australian, n (%)	22 (21.0)	101 (20.0)	OR = 1.02 (95% CI = 0.61–1.71)
Unemployment, n (%)	82 (79.6)	165 (38.0)	OR = 6.35 (95% CI = 3.79–10.65)**
<Year 10 education, n (%)	45 (42.9)	85 (17.6)	OR = 3.53 (95% CI = 2.25–5.55)**

*p < .05; **p < .001. Note. *p < .05; **p < .001. CI = confidence interval; OR = odds ratio. Age range = 18–83 years.

Key mental health and offending characteristics for both the cases and the controls are presented in Table 3. Compared to the controls, the participants with psychosis were significantly more likely to report comorbid lifetime diagnosis of non-psychotic mental disorders including major depression (OR = 4.20, 95% CI = 2.72–6.61, p < .001) and attentional deficit hyperactivity disorder (ADHD; OR = 3.42, 95% CI = 2.00–5.84, p < .001) but less likely to report comorbid anxiety disorders (OR = 5.61, 95% CI = 3.57–8.81, p < .001). A self-harm history (including past suicide attempts) was significantly more likely in those with psychosis (OR = 7.62, 95% CI = 4.81–12.09, p < .001) along with a history of head injury (OR = 2.60, 95% CI = 1.69–3.99, p < .001). Previous illicit substance use was ubiquitous among the participants in both groups, although those with psychosis significantly preferred amphetamines (OR = 4.80, 95% CI = 2.34–9.82, p < .001) and cannabis (OR = 3.40, 95% CI = 1.59–7.28, p = .002). The groups were not found to differ in terms of criminal offending characteristics.

Table 3.

Univariate and multivariate analyses for key mental health, substance use and offending characteristics for psychosis-spectrum cases (n = 105) and controls (unweighted n = 471, weighted = 489).

Variable	Psychosis-spectrum cases, n (%)	Non-psychotic controls, n (%)	Unadjusted analyses	Adjusted analysesa
Lifetime depression	65 (62.5)	135 (28.2)	OR = 4.24 (95% CI = 2.72–6.61)**	OR = 5.57 (95% CI = 3.18–9.76)**
Lifetime anxiety	49 (46.7)	398 (83.1)	OR = 5.61 (95% CI = 3.57–8.81)**	OR = 7.68 (95% CI = 4.25–13.89)**
Lifetime ADHD/ADD	27 (26.0)	45 (9.4)	OR = 3.42 (95% CI = 2.00–5.84)**	OR = 3.14 (95% CI = 1.60–6.18)*
Lifetime other MI	10 (9.5)	52 (10.9)	OR = 0.87 (95% CI = 0.43–1.78)	OR =1.31 (95% CI = 0.53–3.19)
Previous self-harm	61 (58.1)	72 (15.3)	OR = 7.62 (95% CI = 4.81–12.09)**	OR =7.38 (95% CI = 4.13–13.18)**
Lifetime head injury	53 (50.5)	136 (28.3)	OR = 2.59 (95% CI = 1.69–3.99)**	OR = 2.27 (95% CI = 1.33–3.86)*
Amphetamine use	94 (91.3)	267 (68.5)	OR = 4.80 (95% CI = 2.34–9.82)**	OR = 4.84 (95% CI = 2.07–11.33)**
Cannabis use	95 (92.2)	303 (77.7)	OR = 3.40 (95% CI = 1.59–7.28)*	OR = 3.91 (95% CI = 1.60–9.57)*
Intoxicated at offence	62 (60.2)	253 (55.5)	OR = 1.21 (95% CI = 0.78–1.87)	OR = 0.98 (95% CI = 0.57–1.67)
Violent index offence	47 (45.2)	193 (43.8)	OR = 1.06 (95% CI = 0.69–1.63)	OR = 1.38 (95% CI = 0.79–2.42)
Any juvenile detention	35 (33.7)	149 (30.6)	OR = 1.15 (95% CI = 0.74–1.81)	OR = 0.85 (95% CI = 0.48–1.49)

Note. *p < .05; **p < .001; ^aadjusted for previous custody, remand status, relationship status, educational qualification and employment status. ADHD/ADD = attentional deficit hyperactivity disorder/attentional deficit disorder; CI = confidence interval; MI = mental illness; OR = odds ratio.

Even when adjusted for those sociodemographic factors identified as significant on univariate analysis – namely previous custody, remand status, relationship status, educational qualifications and employment status – the key mental health and offending factors that were found to be significant in the unadjusted analyses continued to remain significant (Table 3).

Discussion

The present study aimed to identify individuals with psychosis, as well as those who are at-risk of psychosis, amongst a sample of adult male prisoners referred to prison mental health services following reception screening in NSW and then compare this group to a control sample of non-psychotic prisoners. The key findings indicate that different stages of psychosis are identifiable amongst prisoners referred to prison mental health services – including the UHR, FEP and established psychosis stages – and that subgroups defined by the level of current illness symptomatology can be further identified for both violent and non-violent offenders. These findings support the notion that there is an increased risk of any criminal offending resulting in contact with the criminal justice system across the clinical stages of psychosis, including in the earlier phases of illness. In the present study, the vast majority of prisoners who were referred to prison mental health services for probable psychosis or at-risk features were in the established psychosis stage, and almost two in five remained psychotic at the time of their interview. One in ten reported no history of psychosis, with an equal proportion of these meeting the UHR and FEP criteria. Compared to the prisoners without psychosis, those on the psychosis spectrum exhibited evidence of greater social disadvantage and extensive histories of non-psychotic mental health and substance use problems, highlighting the complexity of their health and other psychosocial needs.

Prevalence of the ARMS and FEP stages among the psychosis-spectrum participants

The prevalence of UHR at 5.7% in the present study is similar to the 5% rate found in a sample of adult male prisoners in London (Jarrett et al., 2012; Jarrett, Valmaggia et al., 2016) but much lower than the 23% rate found in Irish adolescent detainees (Flynn et al., 2012). Sampling processes differed across the studies, with the London study employing a two-stage study-specific screening process (Jarrett et al., 2012; Jarrett, Valmaggia et al., 2016) and the present study relying on the standard reception screening employed by custodial health services. The much higher rate of UHR identified in the Irish study may be, in part, due to the younger age of the participants detained in a juvenile institution and differences in the initial screening processes across the studies (Flynn et al., 2012).

In the present study, the inclusion of individuals meeting UHR criteria may have been limited by the inherent difficulty of detecting the non-specific and attenuated symptoms of UHR in a health service context (Lin et al., 2013). Referrals to in-reach mental health services are likely to be dominated by more overtly unwell prisoners in a clinical stage further along the psychosis spectrum. Given that approximately one in five individuals with UHR are likely to transition to a threshold-level psychotic illness in an average year (Fusar-Poli et al., 2012), and that this rate may be even higher among prisoners due to the stress of imprisonment and the increased levels of disadvantage, providing staff resources and training to prison mental health clinicians that is sufficient for successfully identifying prisoners at-risk for developing psychosis would be a worthwhile intervention, especially in a custodial setting.

The prevalence of FEP at 5.7% in the present study is twice that identified in the London study (Jarrett et al., 2012; Jarrett, Valmaggia et al., 2016), likely reflecting key differences in the two samples (i.e. a mental health referred sample in the present study compared to one in which all prisoners were screened at reception), although this is not reflected in the difference in prevalence of UHR identified in the two studies. Early identification and treatment for the FEP is the gold standard approach in community settings, supported by evidence that a longer duration of untreated psychosis (DUP) is associated with a worse illness prognosis (Marshall et al., 2005; Perkins et al., 2005). Just as with identifying prisoners meeting UHR criteria, there may be a need for additional resources and staff training to ensure that individuals experiencing the FEP are identified as early as possible and offered timely treatment in prison settings. In the context of inadequate resourcing for prison mental health services, clinicians understandably focus on the most overtly mentally unwell, possibly at the expense of early intervention opportunities.

Characteristics across the psychosis-spectrum groups defined by current symptom level

Few differences in sociodemographic, clinical and offending characteristics were found across the psychosis spectrum when participants were grouped according to their current level of psychotic symptoms (i.e. no symptoms, subthreshold symptoms or active psychotic symptoms). The relatively small size of the three groups hampers comparative analysis, but it is also possible that grouping based on current symptom level does not usefully delineate subgroups amongst prisoners on the psychosis spectrum. A larger study able to compare the five clinical stages of psychosis (i.e. UHR, FEP, remitted, residual and chronically psychotic) may enable differences to be identified.

When the relative timings of mental health and criminal justice events were established for the psychosis-spectrum sample as a whole, the onset of possible ARMS experiences, illicit substance use and first criminal justice contact were reported to have occurred around the same age, which on average was between 16 and 17 years, representing a heightened risk period for the emergence of these indicators of complex vulnerability and a key window for early intervention. In many instances, however, criminal offending occurred prior to first contact with psychiatric services among patients with schizophrenia, thus limiting opportunities for health services to successfully intervene (Munkner et al., 2003; Wallace et al., 2004). Similarly, the psychosis-spectrum participants of the present study only had their first contact with mental health services, on average, 4.5 years after the first indicators of psychosis risk, but by then the first criminal charge was likely to have occurred. Furthermore, illness prognosis is typically worsened by a prolonged DUP, such as is seen in the present sample, compared to the typical mean DUP of 1 year observed in community samples in high-income countries such as Australia (Large et al., 2008). This again highlights the need for early intervention targeting mental health and criminogenic needs in this highly disadvantaged group.

Comparing participants with and without psychosis

The psychosis-spectrum participants were more likely to have experienced social adversity (e.g. unemployment, low educational attainment, prior incarceration) and a wide range of non-psychotic mental health problems (e.g. mood and anxiety disorders, self-harm history, illicit drug use history) compared to those without psychosis. The mental health differences persisted even after significant sociodemographic covariates were included in the multivariate analyses, suggesting that poor clinical outcomes amongst prisoners with psychosis-spectrum illness may not be entirely explained by social adversity. In the only previous study on early psychosis amongst adult male prisoners, those with UHR and FEP were found to have higher levels of mental health symptoms and distress than those with other mental disorders (Jarrett, Valmaggia et al., 2016). Nevertheless, there is a possibility that the comorbid mental health symptoms identified among psychosis-spectrum individuals reported in past research as well as in the present study could be attributed to a continuum of a separate illness psychopathology altogether. For example, comorbid depressive symptoms could indicate an underlying mood disorder with psychotic features.

Given the well-established association between psychosis and violent offending, the finding that those on the psychosis spectrum were not more likely to have committed a violent index offence than those without psychosis was unexpected. This contradicts previous research indicating that illness-related factors play a greater role in the aetiology of more severe forms of offending, particularly violent offending (Angermeyer, 2000; Large & Nielssen, 2011; Nielssen et al., 2012; Swanson et al., 2006; Yee et al., 2011), although it may be that this is the case predominantly for homicide offences (Fazel et al., 2009). However, as self-report of diagnoses and symptoms was relied on in order to form the control group of the present study, some individuals without psychosis could have been excluded and some individuals with psychosis could have been included, thus diminishing potential group differences further. Although this method was not completely rigorous, we were limited to the original NPHS data available. Also, while it was not possible to ascertain the specific nature of the violent crime for all of the participants in the present study, another potential explanation may lie in the ubiquitous presence of substance use problems across the groups, which may have dominated any underlying association between psychosis and risk of violent criminal offending (Fazel et al., 2009; Van Dorn et al., 2012).

Strengths and limitations

The present study is the first to identify, categorise and describe prisoners across the full psychosis spectrum – from the ARMS and FEP through to established and chronic psychosis – in a custodial setting. Past research examining the association between psychosis and criminal offending has largely ignored the illness stage (Fazel et al., 2016; Fazel & Seewald, 2012; Hodgins, 2008; Munkner et al., 2003; Wallace et al., 2004). The only previous study on early psychosis (UHR and FEP) amongst adult prisoners does not report on the occurrence of psychosis further along the continuum (Jarrett et al., 2012; Jarrett, Valmaggia et al., 2016). The other key strengths of the present study include the comparison of psychosis-spectrum cases with a non-psychotic prison control group and the use of structured and validated instruments – including the CAARMS (Yung et al., 2005) and the SCID-II (First et al., 1997) – to assess for at-risk symptoms of psychosis and establish subgroups across the psychosis spectrum. Although a validated substance use questionnaire was not employed, participants were asked about their prior history of substance use (including age of onset), allowing for the temporal mapping of this key clinical factor which has been repeatedly implicated in the relationship between psychosis and criminality in relation to onset of both illness and offending.

A key limitation of this study is the selection of cases from a sample of mentally unwell prisoners referred to prison mental health services for possible psychosis or psychosis risk, rather than from an unselected prison population. This likely resulted in an over-representation of established psychosis cases (as well as an under-representation of UHR and FEP cases) and limited the comparative analyses that could be conducted across the psychosis-spectrum groups. Furthermore, we did not interview a random sample of those not referred to prison mental health services and we excluded individuals who were too mentally unwell to consent, too risky to be seen, unable to speak sufficient English or found to have significant cognitive impairment, and this would also likely have resulted in an underestimation of psychosis prevalence. Similarly, the near ubiquitous use of substances prior to incarceration, especially amphetamines in the Australian context (Australian Institute of Health and Welfare, 2020; Nielssen & Misrachi, 2005), would have contributed to the occurrence of both subthreshold and threshold-level psychosis among vulnerable individuals assessed soon after reception to prison. The study also only examined adult prisoners and did not include adolescents, which likely reduced the amount of participants at the early stages of illness. Furthermore, given the refusal rate of 33% among the eligible individuals approached, differences are also expected between participants and non-participants – although a participation rate as low as 67% is to be expected given the difficulties associated with recruiting within a prison setting from among a population of individuals with psychosis.

All of these limitations are likely to have led to an underestimation of psychosis-spectrum prevalence, particularly of those in the earliest clinical stages of illness. Future studies could examine the full psychosis spectrum in a larger, unselected sample of prisoners in order to further establish the prevalence of the various stages of psychosis in this population, including whether or not different stages of illness are associated with different types of offence and whether or not the relationship between psychosis and criminal offending varies along a gradient of illness and offence severity.

Conclusion

The present study is the first to examine the full psychosis spectrum including UHR and FEP as well as established psychosis stages in a prison population. The prisoners on the psychosis spectrum were found to have higher levels of social disadvantage and comorbid mental health and substance use problems, but psychosis was not found to be associated with a greater risk of violent offending. Early intervention services should consider how to more effectively identify those individuals presenting with the ARMS or early psychosis and intervene in order to reduce their risk of criminal justice system contact.

Funding Statement

This work was supported by a fellowship from the New South Wales Institute of Psychiatry (NY), funding (participant payments) and in-kind support from Justice Health and Forensic Mental Health Network, and salary support from the NHMRC Centre for Research Excellence in Offender Health (PC).

Ethical standards

Declaration of conflicts of interest

Natalia Yee has declared no conflicts of interest.

Prabin Chemjong has declared no conflicts of interest.

Daria Korobanova has declared no conflicts of interest.

Suki Scade has declared no conflicts of interest.

Matthew Large has declared no conflicts of interest.

Olav Nielssen has declared no conflicts of interest.

Vaughan Carr has declared no conflicts of interest.

Kimberlie Dean has declared no conflicts of interest.

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the Justice Health and Forensic Mental Health Network Human Research Ethics Committee, the Aboriginal Health & Medical Research Council Human Research Ethics Committee and the Corrective Services New South Wales Ethics Committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed consent

Informed consent was obtained from all individual participants included in the study.